Wojtek Zajdel

Visual surveillance in wide areas (e.g. airports) relies on cameras that observe non-overlapping scenes. Multi-person tracking requires re-identification of people, when they leave one field of view and later enter another. For this we use appearance cues. Under the assumption that all observations of a single person are Gaussian distributed, the o...

Visual surveillance in wide areas (e.g. airports) relies on cameras that observe non-overlapping scenes. Multi-person tracking requires re-identification of a person when he/she leaves one field of view, and later appears at another. For this, we use appearance cues. Under the assumption that all observations of a single person are Gaussian distrib...

This paper presents a smart surveillance system named CASSANDRA, aimed at detecting instances of aggressive human behavior in public environments. A distinguishing aspect of CASSANDRA is the exploitation of the complimentary nature of audio and video sensing to disambiguate scene activity in real-life, noisy and dynamic environments. At the lower l...

This paper presents a surveillance system for tracking mul- tiple people through a wide area with sparsely distributed cameras. The computational core of the system is an adaptive probabilistic model for reasoning about peoples' appearances, locations and identities. The sys- tem consists of two processing levels. At the low-level, individual perso...

In this paper we present Dynamic Bayesian Networks (DBN) as computational framework for the analysis of dynamic systems. We rst give a short overview of the work on state estimation and system identication. Then we present two application where the DBN is used: robot localization and the tracking of multiple persons with multiple camera's.

Robots are conveniently controlled by a human operator with spoken commands, since voice is a natural communication medium for humans. In order to successfully carry out a command, a robot needs to know which of the possibly many people gave the command and where this person is located. In this paper, we present a particle-filter based algorithm fo...

We consider a visual scene analysis scenario where objects (e.g. people, cars) pass through the viewing field of a static camera and need to be detected and segmented from the background. For this purpose, we introduce a hybrid dynamic Bayesian network and derive an expectation propagation (EP) algorithm for robust estimation of object shapes and a...

In this paper we describe a system that enables a mobile robot equipped with a color vision system to track humans in indoor environments. We developed a method for tracking humans when they are within the field of view of the camera, based on motion and color cues. However, the robot also has to keep track of humans which leave the field of view a...

The paper presents a novel method for online tracking of multiple objects with non-overlapping cameras. The method is based on a generative model defining probabilistic dependencies between observations, the underlying color properties of objects and their dynamics. It allows for a full Bayesian inference of trajectories. We developed an online alg...

The paper presents a novel method for online tracking of multiple objects with non-overlapping cameras. The method is based on a generative model defining probabilistic dependencies between observations, the underlying color properties of objects and their dynamics. It allows for a full Bayesian inference of trajectories. We developed an on-line al...

In a human-inhabited environment it is essential that a robot which interacts with humans is able to keep track of them when they move around in the environment. This is not an easy job. Multiple people may be in the robots vicinity, sometimes a person leaves the vicinity of the robot and reenters some time later. Two issues are essential in tracki...

Tracking with multiple cameras requires partitioning of observations from various sensors into trajectories. In this paper we assume that the observations are generated by a hidden, stochastic 'partition' process and propose a hidden Markov model (HMM) as a generative model for the data. The state space for the hidden variable is intractable, so th...

For fully automating surveillance, or in general, monitoring tasks, it is often required to be able to follow a person or object over multiple cameras. To do so, these objects have to be recognized and identified on differing backgrounds, lighting conditions and viewing angles. This thesis describes a method of gathering information about the appea...

Tracking with multiple cameras requires partitioning of ob-servations from various sensors into trajectories. In this paper we assume that the observations are generated by a hidden, stochastic 'partition' process and propose a hidden Markov model (HMM) as a generative model for the data. The state space for the hidden variable is intractable, so t...

We present an algorithm for tracking many objects observed with dis- tributed, non-overlapping sensors. Our method is derived from a proposi- tion that the observations of some constant, intrinsic properties of an object form a cluster (eg. in the color space). However sensors also provide dynamic data about an object like time and location. Tracki...

For a flexible camera-to-camera tracking of multiple objects we model the object's behavior with a Bayesian network and combine it with the multiple hypothesis framework that associates observations with objects. Bayesian networks offer a possibility to factor complex, joint distributions into a product of intuitive conditional densities describing...

We present a probabilistic method for online tracking of multiple objects with sparsely distributed cameras. The method explicitly identifles objects with latent labels and introduces latent states that represent intrinsic appearance properties. The dependency between observations, discrete labels and continuous states leads to a hybrid model, whic...

Proefschrift Universiteit van Amsterdam. Met lit. opg. - Met samenvatting in het Nederlands. Ook verschenen als on line resource.